基于GPU的邻近粒子搜索优化算法

Improved neighbor particle search algorithm based on GPU

为提高邻近粒子的搜索速度,基于硬件图形处理器(GPU),采用SPH方法中交替使用邻域区域相关搜索法和条形Point-In-Box(PIB)搜索法的机制,实现一种基于GPU的邻近粒子融合搜索算法。利用该方法对典型的流体溃坝现象进行模拟,在搜索效率和搜索准确率方面,与基于CPU的融合搜索法和基于GPU的多维树搜索法进行对比。实验结果表明,该算法的性能随着粒子数的增多,表现更优。

To improve the search speed of neighboring particles, a merge search algorithm based on graphic processing unit （GPU）,in which neighborhood correlation search algorithm and striped Point-In-Box （PIB） search algorithm worked alterna-tively on GPU, was employed. The typical dam break phenomenon was simulated using SPH with merge search algorithm based on GPU. And for search efficiency and accuracy, the proposed algorithm, merge search algorithm based on CPU and multidi-mensional tree searching algorithm on GPU were compared. Experimental results indicate that the performance of the improved algorithm gets better with the increase of the number of particles.